VALUE IN HEALTH 17 (2014) 288–296

Available online at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/jval

SYSTEMATIC REVIEW

Socioeconomic Status and Nonadherence to Antihypertensive Drugs: A Systematic Review and Meta-Analysis M.H.D. Wasem Alsabbagh, BSc1, Mark Lemstra, PhD1, Dean Eurich, PhD2, Lisa M. Lix, PhD3, Thomas W. Wilson, MD4, Erin Watson, MLIS5, David F. Blackburn, PharmD1,* 1 College of Pharmacy and Nutrition, University of Saskatchewan, Saskatoon, SK, Canada; 2Department of Public Health Sciences, School of Public Health, University of Alberta, Edmonton, AB, Canada; 3Faculty of Medicine, Department of Community Health Sciences, University of Manitoba, Winnipeg, MB, Canada; 4Department of Medicine, Royal University Hospital, Saskatoon Health Region, Saskatoon, SK, Canada; 5Health Sciences Library, University of Saskatchewan, Saskatoon, SK, Canada

AB STR A CT

Background: Although conventional wisdom suggests that low socioeconomic status (SES) is a robust predictor of medication nonadherence, the strength of this association remains unclear. Objectives: 1) To estimate the proportion of studies that identified SES as a potential risk indicator of nonadherence, 2) to describe the type of SES measurements, and 3) to quantify the association between SES and nonadherence to antihypertensive pharmacotherapy. Methods: A systematic review and meta-analysis research design was used. We searched multiple electronic databases for studies in English or French examining nonadherence to antihypertensive medications measured by electronic prescription databases where explanatory factors were considered. Two authors independently assessed quality, described the SES measure(s), and recorded its association with nonadherence to antihypertensives. A random-effects model metaanalysis was performed, and heterogeneity was examined by using the I2 statistic. Results: Fifty-six studies with 4,780,293 subjects met the inclusion criteria. Twenty-four of these studies (43%) did not report any SES measures. When it was reported (n ¼ 32), only seven

(13%) examined more than one component but none performed a multidimensional assessment. Most of the studies relied on income or income-related measures (such as prescription-drug benefits or copayments) (27 of 32 [84%]). Meta-analysis could be quantified in 40 cohorts reported in 30 studies. Overall, the pooled adjusted risk estimate for nonadherence according to SES (high vs. low) was 0.89 (95% confidence interval 0.87–0.92; I2 ¼ 95%; P o 0.001). Similar patterns were observed in all subgroups examined. Conclusions: Published studies have not found a strong association between low SES and nonadherence to antihypertensive medications. However, important limitations in the assessment of SES can be identified in virtually all studies. Future studies are required to ascertain whether a stronger association is observed when SES is determined by comprehensive measures. Keywords: antihypertensive medications, drug adherence, pharmacy, population-level, socioeconomic status.

Introduction

these factors may affect regular medication use [8]. Although the relationship between SES and nonadherence has been inconsistent [9], we hypothesized that methodological approaches may have attenuated an important relationship. Electronic prescription databases are the methods most frequently used for the assessment of nonadherence [10,11]; however, they often lack important patientlevel information required to describe SES in detail. As a result, indirect measures of income such as receipt of prescription-drug benefits through government co-payments are often used as sole indicators of SES in many studies [12–15], while direct measures of income from taxation records are rarely used [6]. In addition, studies generally do not account for SES factors not related to income and even fewer incorporate multiple SES measurements representing different dimensions [16,17]. Finally, several population-based studies can be identified in which SES factors are absent altogether [18–20].

Socioeconomic status (SES) is a multidimensional construct that represents an individual’s position relative to other people in the community. It is commonly considered a product of the interaction between material and social factors. Material factors include income, education, and employment, whereas social factors are usually represented by living arrangements and family structure [1]. In health care research, low SES has proven to be a strong predictor of health care utilization, morbidity, and premature death [2–5]. Nonadherence to chronic medications, such as antihypertensives (AHTs), can also be determined by low SES [6,7]. SES is an intriguing factor in the search for determinants of population-level nonadherence to AHTs because of its associations with economic, social, and education-related factors. Indeed, all

Copyright & 2014, International Society for Pharmacoeconomics and Outcomes Research (ISPOR). Published by Elsevier Inc.

* Address correspondence to: David F. Blackburn, College of Pharmacy and Nutrition, University of Saskatchewan, 110 Science Place, Saskatoon, SK, Canada S7N 5C9. E-mail: [email protected]. 1098-3015/$36.00 – see front matter Copyright & 2014, International Society for Pharmacoeconomics and Outcomes Research (ISPOR). Published by Elsevier Inc. http://dx.doi.org/10.1016/j.jval.2013.11.011

VALUE IN HEALTH 17 (2014) 288–296

To understand the extent to which SES may affect nonadherence to AHT medications, as well as the approaches used to account for it, we conducted a systematic review and meta-analysis of the literature pertaining to SES and nonadherence to AHT medications by using population-based electronic prescription data. Our study had three objectives: 1) to estimate the proportion of studies that identified SES as a potential risk indicator of nonadherence, 2) to describe the type of SES measurements that were used in each study, and 3) to quantify the association between SES and nonadherence to AHT medications.

289

corresponding to the measure of SES. In addition, we abstracted information on region of origin, publication year target medication(s), adherence measurement (medication possession ratio [MPR]-related vs. discontinuation-related [11]), number of subjects (total/low SES/ higher SES), follow-up (observation) days, and the number of SES domains captured in each study. We categorized SES covariates as income-related and non–income-related and subcategorized incomerelated SES covariates as follows: income level, health-plan coverage or medication co-payment amount, and receipt of social assistance or income security benefits. Income-level factors were identified as direct (e.g., by linking taxation data to dispensation records) or indirect (e.g., from census neighborhood or coverage type).

Methods Search Strategy We used a comprehensive search strategy of electronic databases including Medline (OVID, 1964 to February 24, 2012), Embase (OVID, 1947 to February 24, 2012), International Pharmaceutical Abstracts (OVID, 1970 to January 31, 2012), the Cochrane Library (Wiley, 1800 to February 28, 2012), Scopus (Elsevier, from 1823 to February 28, 2012), CINAHL (EBSCO, 1937 to February 28, 2012), PsycINFO (OVID, 1806 to February week 3, 2012), Sociological Abstracts (ProQuest, 1952 to February 28, 2012), ProQuest Dissertations & Theses (1639 to February 28, 2012), Web of Science (Thomson Reuters, 1899 to February 28, 2012), and OAIster (1975 to Dec 31, 2011). Reference lists of identified articles were manually searched for additional studies not captured in the electronic database reviews.

Study Selection Studies were included if they satisfied the following criteria: 1) examined nonadherence to AHT medications; 2) used electronic prescription databases as the source of nonadherence information; 3) conducted multivariable modeling to determine the independent effect of explanatory covariates on the outcome of nonadherence; and 4) were published in English or French. Eligible AHT medications included angiotensin-converting enzyme inhibitors or angiotensin receptor blockers, beta-blockers, calcium channel blockers, or thiazide diuretics for any indication. Studies were not restricted by design or publication date. In some cases, we contacted authors of studies to clarify or obtain data.

Review Procedure and Assessment of Methodological Quality The identification of studies was carried out in two steps. First, after removal of duplicates, two of the authors (M.W.A. and M.L.) examined the titles and abstracts identified in the initial search. Second, the same two reviewers (M.W.A. and M.L.) examined fulltext articles for each study identified in the first step for both eligibility and methodological quality. Disagreement between the two reviewers was resolved by additional review and discussion and then, if required, with tie-breaking by a third reviewer (D.F. B.). Quality of included studies was assessed by the reviewers (M. W.A. and M.L.) with a checklist developed by the International Society for Pharmacoeconomics and Outcome Research (ISPOR) for retrospective database studies [21,22]. This checklist has been used in systematic reviews previously [23], and it consists of 27 quality review questions related to data source, research design, study population, variable definitions, statistics, and discussion. From each study, we determined whether a SES measure was assessed by identifying any material factor (e.g., income, education, or employment) or social factor (e.g., living alone, or family structure) that was included in the nonadherence model. In cases in which these traditional SES measures were not used, we identified ethnicity as a possible indirect measure in a sensitivity analysis. We also recorded estimates of the ratio effects measures of nonadherence to AHT medications along with 95% confidence intervals (95% CIs)

Statistical Analysis We assessed the agreement in study inclusion/exclusion between reviewers in each step by using Cohen’s kappa statistic (κ) [24]. We adopted the following interpretations for κ: 0.60 o κ was considered low agreement, 0.60 r κ r 0.79 was considered good agreement, and κ Z 0.80 was considered very good agreement [25]. We assessed heterogeneity by using the I2 statistic and the corresponding Tau-squared (τ2) test. This statistic represents the proportion of variability that can be attributed to between-studies variability [26]. We adopted the following interpretations for the I2 statistic: 0% to 40%, low heterogeneity; 41% to 74%, moderate heterogeneity; and 75% to 100%, considerable heterogeneity. All estimates were pooled where possible irrespective of the level of heterogeneity observed and subgroup analyses completed to explore potential sources of study heterogeneity [27]. We then conducted a random-effects model meta-analysis by using the inverse-variance method to estimate the effect of SES on medication nonadherence from the pooled data [26–28]. A random-effects model accounts for potential heterogeneity between the populations and unmeasured confounding [26]. For studies reporting more than one nonadherence measure, we prioritized MPR-related outcomes over other measures. The MPR is calculated usually by summing all days’ supply during a certain period of observation of the medication and dividing it by the total days of that period [10]. For studies reporting more than one SES variable, we used the measure with the largest effect size regardless of the direction of the association. When categorical SES measures had more than two levels, we reported the risk estimate of the highest SES level relative to the lowest. Subgroup analyses were conducted for type of nonadherence measurement, type of medication, type of SES measurement, and region of origin of the data. We used the Z test for overall effects and the chi-square statistic to test for differences in betweengroups effects [29]. Finally, because ethnicity may be considered an indirect SES measurement [30,31], we performed sensitivity analyses by including ethnicity as a measure of SES to assess the proportion of studies that identified SES as a potential risk indicator of nonadherence. In addition, we performed a sensitivity analysis by using the estimate with the lowest effect size instead of the highest for studies reporting more than one SES variable. We evaluated publication bias visually by using the funnel plot [32]. We adopted the protocol developed by The Cochrane Collaboration and used Review Manager (Version 5.1.7, Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2011) to perform the meta-analysis [27].

Results Included Studies Our search identified 11,351 titles/abstracts, with 56 studies meeting our inclusion criteria (Fig.1; eTable 1). Overall agreement for inclusion/exclusion of studies between the reviewers was

290

VALUE IN HEALTH 17 (2014) 288–296

Fig. 1 – Flow chart for titles/abstracts and articles included in the review. CI, confidence interval. found to be good (κ ¼ 0.79, and κ ¼ 0.69 for first and second steps, respectively) (Fig. 1). Of these 56 included studies, 8 studies [33– 39] scored less than 50% on the methodological quality review. The median methodological quality score was 68%, and the interquartile range was 19%. Table 1 summarizes the characteristics of included studies. The total number of subjects included in our review was 4,708,293 (range 236–1,075,285 per study). Most of the studies were conducted in Europe (11 of 56 [20%]) and North America (34 of 56 [61%]). Studies ranged from assessing one AHT medication (6 of 56 [11%]) to assessing combinations of two medications or more. Nonadherence was measured by discontinuation/nonpersistence (25 of 56 [45%]), MPR-related measurements (26 of 56 [47%]), or both (5 of 56 [8%]). The follow-up duration was 180 days or less in 13 studies, 181 days to 365 days in 31 studies, and more than 365 days in 8 studies.

SES Measures Overall, 24 of 56 studies (43%) did not assess SES with any material or social measure. When ethnicity was considered as

an eligible SES measure, however, the proportion of studies with no SES measure decreased to 19 of 56 studies (34%). An SES measure was lacking in all studies published before 2004 (9 studies), compared with 15 of 47 studies (32%) published between 2004 and 2012. Of the 32 studies that assessed SES, 2 studies did not report estimates of the effect of SES on nonadherence and could not be obtained from authors [40,41]. No study organized SES variables into multidimensional scales or indices such as a deprivation index [42,43]. From all studies that assessed SES, 25 studies (78% [25 of 32]) identified only one SES measure among their study subjects. Of these, prescription-drug coverage or medication co-payment amount based on income was most commonly used (17 of 25 studies) [9,15,38,44–57]. Four studies captured the income level [41,58–60], three studies used social assistance benefits or income security benefits [61–63], and only one study used the education level [64]. Seven studies (22% [7 of 32]) identified more than one SES measure in their study population. All the studies contained at least one income-related variable, and three identified incomerelated measures only [36,39,65]. Nonincome variables in the

291

VALUE IN HEALTH 17 (2014) 288–296

Table 1 – Characteristics of included studies. Characteristic

Region of data origin North American Europe Other countries Date of publication Before 2004 2004 and after AHT medication(s) studied ACEIs ARBs BBs CCBs HCTZ Other (if more than one medication was studied) Adherence measure type MPR related Discontinuation Both Number of SES measurements None 1 41 If SES was not assessed (N ¼ 24), did study assess ethnicity? Yes No SES measure One SES measure not related to income Education One SES measure related to income Prescription drug coverage or medication co-payment amount Income level Social assistance benefits or income security benefits More than one SES measure Two income-related measures Three income-related measures Income-related measure þ one measure not related to income Income-related measure þ two measures not related to income Method to measure income Direct Link with taxation records Indirect Health plan Neighborhood Follow-up days category Up to 180 181–365 4365 –

Number of subjects

% of total sample (n ¼ 4,708,293)

Number of studies (%) (N ¼ 56)

2,666,720 880,224 1,161,349

56.6 18.7 24.7

35 (62.5) 10 (17.9) 11 (19.6)

206,025 4,502,268

4.4 95.6

9 (16.1) 47 (83.9)

27,114 18,396 102,690 29,324 17,949 4,512,820

0.6 0.4 2.2 0.6 0.4 95.8

6 2 6 2 2 38

3,058,762 1,351,749 297,782

65 28 6.3

26 (46.4) 25 (44.6) 5 (9.0)

2,961,112 1,546,235 200,946

62.9 32.8 4.3

24 (42.9) 25 (44.6) 7 (12.5)

1,761,764 1,199,348

37.4 25.5

5 (8.9) 19 (33.9)

28,395

0.6

1 (1.8)

1,171,195 267,379 79,266

24.9 5.7 1.7

17 (30.4) 4 (7.1) 3 (5.4)

4,525 24,443 104,200 14,219

0.1 0.5 2.2 0.3

2 1 3 1

150,609

3.2

3 (5.4)

53,603 200,370

1.1 4.3

2 (3.6) 2 (3.6)

1,426,148 2,753,990 502,138 26,107

30.3 58.5 10.7 0.6

13 31 8 4

(10.7) (3.6) (10.7) (3.6) (3.6) (67.8)

(3.6) (1.8) (5.4) (1.8)

(23.2) (55.4) (14.3) (7.1)

ACEIs, angiotensin-converting enzyme inhibitors; AHT, antihypertensive; ARBs, angiotensin receptor blockers; BBs, beta-blockers; CCBs, calcium channel blockers; HCTZ, hydrochlorothiazide; MPR, medication possession ratio.

remaining studies included education and employment [66], education [67], living alone [40], or household composition [6]. In total, seven studies captured the income level either alone or in addition to other SES measures. Of these, classification into income groups was obtained directly by linking taxation data to dispensation records in three studies [6,59,67], indirectly from census neighborhood income in two studies [40,58], and indirectly from low-income drug coverage in two studies [60–65].

Nonadherence with Higher SES Excluding the two studies in which an estimate could not be obtained [40,41], we extracted data for 40 cohorts in 30 studies reporting an SES variable. Higher SES was associated with a lower risk of nonadherence in 31 of 40 cohorts (77.5%), with no difference in 1 cohort, and with a higher risk of nonadherence in 8 cohorts (Fig. 2). Overall, the pooled adjusted risk estimate indicated a lower risk of nonadherence among individuals with a higher SES: 0.89 (95% CI 0.87–0.92; P o 0.001); however, high

292

VALUE IN HEALTH 17 (2014) 288–296

heterogeneity was observed (τ ¼ 0.01; I2 ¼ 95%). Inspection of the funnel plot did not suggest potential publication bias (Fig. 3). To explore heterogeneity in the results, we performed several subgroup analyses. Similar results, however, were observed in studies scoring above 50% on the quality checklist (pooled adjusted risk estimate 0.90; 95% CI 0.87–0.92; I2 ¼ 95%) and scoring below 50% (pooled adjusted risk estimate 0.86; 95% CI 0.66–1.12; I2 ¼ 84%). A subgroup analysis was not performed on measures not related to income because they were identified only in one study. Studies that used discontinuation as the end point (pooled adjusted risk estimate 0.91; 95% CI 0.87–0.96; I2 ¼ 92%) showed results consistent with those that used MPR-related measures (0.88; 95% CI 0.85–0.92; I2 ¼ 93%), and studies from

North America and Europe (0.90; 95% CI 0.87–0.94; I2 ¼ 95%) produced results similar to those from other countries (0.86; 95% CI 0.81–0.92; I2 ¼ 87%). In the sensitivity analysis, using the measure with the lowest effect size (for studies reporting more than one SES variable) did not change the pooled adjusted risk estimate (0.90; 95% CI 0.88– 0.92; I2 ¼ 95%). Smaller heterogeneity was observed in certain cohorts restricted by the specific type of medication used. Pooled adjusted risk estimates representing the effect of higher SES were different for cohorts receiving angiotensin-converting enzyme inhibitors (0.83; 95% CI 0.79–0.88; I2 ¼ 0%), beta-blockers (0.77; 95% CI 0.66–0.9; I2 ¼ 95%), calcium channel blockers (0.98; 95% CI 0.85– 1.14; I2 ¼ 70%), hydrochlorothiazide (0.81; 95% CI 0.74–0.90;

Fig. 2 – Pool risk estimates of nonadherence with high socioeconomic status. CI, confidence interval; SE, standard error; SES, socioeconomic status.

VALUE IN HEALTH 17 (2014) 288–296

293

Fig. 3 – Funnel plot of studies identified in the systematic review. SE, standard error.

I2 ¼ 0%), or angiotensin receptor blockers (1.0; 95% CI 1.0–1.0; I2 ¼ 84%), where the test for subgroup difference was significant (P o 0.001) (Fig. 4). However, the pooled estimate remained relatively consistent with the overall findings.

Discussion Among published studies of nonadherence to AHT medications using electronic prescription databases, 43% did not account for any SES measure despite their theorized importance as a determinant of nonadherence [8]. When SES was assessed, the vast majority of studies identified single factors relating to income and none examined SES by using a comprehensive measure. Pooled analyses indicated that higher SES is associated with an 11% decrease in the adjusted risk of nonadherence; however, heterogeneity between included studies was very high. Although the quantitative effect of SES on medication nonadherence cannot be confirmed from this meta-analysis because of high levels of heterogeneity, incomplete (or absent) SES measures, and inconsistent approaches in the literature, these results clearly demonstrate major deficiencies in previous attempts to understand this complex issue. As a result, SES cannot be considered a strong predictor of medication nonadherence because the evidence supporting this view remains theoretical at best, at least with respect to AHT medications. In reality, it must be recognized that current knowledge about SES and nonadherence is extremely poor. To our knowledge, this is the first systematic review specifically evaluating the effects of SES on nonadherence to AHT medications and the results suggest that more research is needed to ensure a consistent and comprehensive approach to the assessment of SES as a possible risk indicator. We chose hypertension as a disease state because of its prevalence, chronicity, lack of symptoms, and treatability [68]. It has been clearly shown that the prevalence of nonadherence to AHT medications is high and adverse health outcomes are commonly observed compared with those demonstrating optimal adherence [69–71]. As a result, even small improvements in nonadherence are likely considered clinically meaningful [72–75]. The most probable explanation for these findings is that administrative databases including electronic prescription

databases do not have ready access to SES information. The importance of the SES factors as potential risk indicators of nonadherence, however, may also be underrecognized. Accordingly, our understanding of the complex relationship between SES and medication nonadherence is likely incomplete because it is based on studies using a very limited set of SES measures, at least in studies focused on AHT medications. SES could be an important determinant of nonadherence through its effect on not only affordability and access to medications but also health literacy and medication knowledge [76]. Indeed, higher SES reduced the risk estimate of nonadherence in 31 of 40 cohorts examined; however, the opposite effect of SES was observed in 8 of 40 cohorts examined. Thus, our systematic review confirms and characterizes the inconsistent findings relating to SES and nonadherence reported in previous narrative reviews [11,77,78]. Our review had several limitations. First, our study examined AHT studies only, and so the results may not generalize to other chronic disease medications. Second, we included only those studies that used electronic prescription databases as the source of nonadherence information. Electronic prescription databases, however, are the most frequently used source of nonadherence information among large populations [10]. Third, it is highly likely that the pooled risk estimates were affected by the lack of detailed SES information and the inconsistent approaches to SES measurement in the published literature. Indeed, the vast majority of SES measures were restricted to income-related measures. Fourth, we used the checklist developed by ISPOR for retrospective database studies to assess publications’ quality. However, a new questionnaire developed by the Academy of Managed Pharmacy/National Pharmaceutical Council/ISPOR Comparative Effectiveness Research Collaborative Initiative could have improved our quality assessment [79]. Last, although we did not observe any publication bias, it is possible that negative studies assessing SES and nonadherence could not be published. SES is frequently overlooked in studies of nonadherence to AHT medications using electronic prescription databases, and it has never been examined in a comprehensive way. Based on the available literature, higher SES appears to be associated with a small reduction in the occurrence of nonadherence to AHT medications. This estimate, however, is based on studies that contained many limitations. Thus, more research is clearly

294

VALUE IN HEALTH 17 (2014) 288–296

Fig. 4 – Pool risk estimates of nonadherence with high socioeconomic status, stratified by medication studied. CI, confidence interval.

VALUE IN HEALTH 17 (2014) 288–296

needed to help clarify this relationship. Considering the public health importance of this outcome and the relative lack of knowledge about its determinants, failure in taking SES into account could prevent targeting of interventions for those who need them. Source of financial support: David F. Blackburn is the Chair in Patient Adherence within the College of Pharmacy and Nutrition, University of Saskatchewan. This position was created through unrestricted financial support from AstraZeneca Canada, Merck Frosst Canada, Pfizer Canada, and the Province of Saskatchewan’s Ministry of Health. Dean Eurich receives salary support though a population health investigator award from the Alberta Heritage Foundation for Medical Research and is a Canadian Institutes of Health Research New Investigator. Lisa M. Lix is supported by a Manitoba Research Chair. M. Wasem Alsabbagh also receives a scholarship through the Research Chair fund. We did not have financial relationships with any organization that might have an interest in the submitted work in the previous 3 years, and no other relationships or activities that could appear to have affected the submitted work. R EF E R EN CE S

[1] Williams D, Moss N, Krieger N. Measuring social class in US public health research: concepts, methodologies, and guidelines. Annu Rev Public Health 1997;18:341–78. [2] Meijer M, Rohl J, Bloomfield K, et al. Do neighborhoods affect individual mortality? A systematic review and meta-analysis of multilevel studies. Soc Sci Med 2012;74:1204–12. [3] Manrique-Garcia E, Sidorchuk A, Hallqvist J, et al. Socioeconomic position and incidence of acute myocardial infarction: a meta-analysis. J Epidemiol Community Health 2011;65:301–9. [4] Smith DF, Boss EF. Racial/ethnic and socioeconomic disparities in the prevalence and treatment of otitis media in children in the United States. Laryngoscope 2010;120:2306–12. [5] Saydah SH, Imperatore G, Beckles GL. Socioeconomic status and mortality: contribution of health care access and psychological distress among U.S. adults with diagnosed diabetes. Diabetes Care 2013;36:49–55. [6] Corrao G, Zambon A, Parodi A, et al. Do socioeconomic disparities affect accessing and keeping antihypertensive drug therapy? Evidence from an Italian population-based study. J Hum Hypertens 2009;23:238–44. [7] Mauskop A, Borden WB. Predictors of statin adherence. Curr Cardiol Rep 2011;13:553–8. [8] Sabaté E. Adherence to Long-Term Therapies: Evidence for Action. Geneva, Switzerland: World Health Organization, 2003. [9] Van Wijk BL, Klungel OH, Heerdink ER, et al. Rate and determinants of 10-year persistence with antihypertensive drugs. J Hypertens 2005;23:2101–7. [10] Andrade SE, Kahler KH, Frech F, et al. Methods for evaluation of medication adherence and persistence using automated databases. Pharmacoepidemiol Drug Saf 2006;15: 565, 74; discussion 575–7. [11] Osterberg L, Blaschke T. Adherence to medication. N Engl J Med 2005;353:487–97. [12] Neugut AI, Subar M, Wilde ET, et al. Association between prescription co-payment amount and compliance with adjuvant hormonal therapy in women with early-stage breast cancer. J Clin Oncol 2011;29:2534–42. [13] Batal HA, Krantz MJ, Dale RA, et al. Impact of prescription size on statin adherence and cholesterol levels. BMC Health Serv Res 2007;7:175. [14] Vinker S, Shani M, Baevsky T, et al. Adherence with statins over 8 years in a usual care setting. Am J Manag Care 2008;14:388–92. [15] Wong MC, Jiang JY, Griffiths SM. Factors associated with compliance to thiazide diuretics among 8551 Chinese patients. J Clin Pharm Ther 2011;36:179–86. [16] Chodick G, Shalev V, Gerber Y, et al. Long-term persistence with statin treatment in a not-for-profit health maintenance organization: a population-based retrospective cohort study in Israel. Clin Ther 2008;30:2167–79. [17] Nahon S, Lahmek P, Saas C, et al. Socioeconomic and psychological factors associated with nonadherence to treatment in inflammatory bowel disease patients: results of the ISSEO survey. Inflamm Bowel Dis 2011;17:1270–6. [18] Avorn J, Monette J, Lacour A, et al. Persistence of use of lipid-lowering medications: a cross-national study. JAMA 1998;279:1458–62.

295

[19] Donnelly LA, Doney AS, Morris AD, et al. Long-term adherence to statin treatment in diabetes. Diabet Med 2008;25:850–5. [20] Caspard H, Chan AK, Walker AM. Compliance with a statin treatment in a usual-care setting: retrospective database analysis over 3 years after treatment initiation in health maintenance organization enrollees with dyslipidemia. Clin Ther 2005;27:1639–46. [21] Peterson AM, Nau DP, Cramer JA, et al. A checklist for medication compliance and persistence studies using retrospective databases. Value Health 2007;10:3–12. [22] Motheral B, Brooks J, Clark MA, et al. A checklist for retrospective database studies—report of the ISPOR Task Force on Retrospective Databases. Value Health 2003;6:90–7. [23] Murphy CC, Bartholomew LK, Carpentier MY, et al. Adherence to adjuvant hormonal therapy among breast cancer survivors in clinical practice: a systematic review. Breast Cancer Res Treat 2012;134:459–78. [24] Cohen J. A coefficient of agreement for nominal scales. Educ Psychol Meas 1960;20:37–46. [25] Douglas GA. Practical Statistics for Medical Research. London, UK: Chapman and Hall, 1991: 403–5. [26] Huedo-Medina TB, Sanchez-Meca J, Marin-Martinez F, et al. Assessing heterogeneity in meta-analysis: Q statistic or I2 index? Psychol Methods 2006;11:193–206. [27] Higgins JPT Green S, eds. Cochrane Handbook for Systematic Reviews of Interventions Chichester, UK: John Wiley & Sons, Ltd., 2006. [28] DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7:177–88. [29] Deeks JJ. Statistical methods for examining heterogeneity and combining results from several studies in meta-analysis. In: Altman DG, Bradburn MJ, eds. Systematic Reviews in Health Care: MetaAnalysis in Context (2nd ed.). London, UK: BMJ Publishing Group, 2008: 285–312. [30] Kaufman JS, Cooper RS, McGee DL. Socioeconomic status and health in blacks and whites: the problem of residual confounding and the resiliency of race. Epidemiology 1997;8:621–8. [31] Oakes JM, Rossi PH. The measurement of SES in health research: current practice and steps toward a new approach. Soc Sci Med 2003;56:769–84. [32] Egger M, Davey Smith G, Schneider M, et al. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997;315:629–34. [33] Charles H, Good CB, Hanusa BH, et al. Racial differences in adherence to cardiac medications. J Natl Med Assoc 2003;95:17–27. [34] Bailey JE, Lee MD, Somes GW, et al. Risk factors for antihypertensive medication refill failure by patients under Medicaid managed care. Clin Ther 1996;18:1252–62. [35] Glader EL, Sjolander M, Eriksson M, et al. Persistent use of secondary preventive drugs declines rapidly during the first 2 years after stroke. Stroke 2010;41:397–401. [36] Zeng F, Patel BV, Andrews L, et al. Adherence and persistence of singlepill ARB/CCB combination therapy compared to multiple-pill ARB/CCB regimens. Curr Med Res Opin 2010;26:2877–87. [37] Brixner DI, Jackson KC II, Sheng X, et al. Assessment of adherence, persistence, and costs among valsartan and hydrochlorothiazide retrospective cohorts in free-and fixed-dose combinations. Curr Med Res Opin 2008;24:2597–607. [38] Patel BV, Remigio-Baker RA, Thiebaud P, et al. Improved persistence and adherence to diuretic fixed-dose combination therapy compared to diuretic monotherapy. BMC Fam Pract 2008;9:61. [39] Liberman JN, Hutchins DS, Shrank WH, et al. Adherence to medication under mandatory and voluntary mail benefit designs. Am J Manag Care 2011;17:e260–9. [40] Khan NA, Yun L, Humphries K, et al. Antihypertensive drug use and adherence after stroke: are there sex differences? Stroke 2010;41:1445–9. [41] Lai EJ, Grubisic M, Palepu A, et al. Cardiac medication prescribing and adherence after acute myocardial infarction in Chinese and South Asian Canadian patients. BMC Cardiovasc Disord 2011;11:56. [42] Carstairs V. Deprivation indices: their interpretation and use in relation to health. J Epidemiol Community Health 1995;49(Suppl. 2):S3–8. [43] Pampalon R, Hamel D, Gamache P, et al. A deprivation index for health planning in Canada. Chronic Dis Can 2009;29:178–91. [44] Wong MC, Su X, Jiang JY, et al. Profiles of discontinuation and switching of thiazide diuretics: a cohort study among 9398 Chinese hypertensive patients. Hypertens Res 2011;34:888–93. [45] Wong MC, Jiang JY, Griffiths SM. Antihypertensive drug adherence among 6408 Chinese patients on angiotensin-converting enzyme inhibitors in Hong Kong: a cohort study. J Clin Pharmacol 2010;50:598–605. [46] Wong MC, Lau RK, Jiang JY, et al. Discontinuation of angiotensinconverting enzyme inhibitors: a cohort study. J Clin Pharm Ther 2012;37:335–41. [47] Wong MC, Jiang JY, Griffiths SM. Adherence to combination therapy among ethnic Chinese patients: a cohort study. Hypertens Res 2010;33:416–21.

296

VALUE IN HEALTH 17 (2014) 288–296

[48] Wong MC, Jiang JY, Su X, et al. Individuals at risk of beta-blocker discontinuation: a cohort study in 19,177 Chinese patients. Clin Res Cardiol 2010;99:277–84. [49] Wong MC, Jiang JY, Griffiths SM. Factors associated with compliance, discontinuation and switching of calcium channel blockers in 20,156 Chinese patients. Am J Hypertens 2009;22:904–10. [50] Wong MC, Jiang JY, Gibbs T, et al. Factors associated with antihypertensive drug discontinuation among Chinese patients: a cohort study. Am J Hypertens 2009;22:802–10. [51] Wong MC, Jiang JY, Griffiths SM. Factors associated with antihypertensive drug compliance in 83,884 Chinese patients: a cohort study. J Epidemiol Community Health 2010;64:895–901. [52] Wong MC, Jiang JY, Griffiths SM. Short-term adherence to beta-blocker therapy among ethnic Chinese patients with hypertension: a cohort study. Clin Ther 2009;31:2170, 7; discussion 2150–1. [53] Kramer JM, Hammill B, Anstrom KJ, et al. National evaluation of adherence to beta-blocker therapy for 1 year after acute myocardial infarction in patients with commercial health insurance. Am Heart J 2006;152(454):e1–8. [54] Sung SK, Lee SG, Lee KS, et al. First-year treatment adherence among outpatients initiating antihypertensive medication in Korea: results of a retrospective claims review. Clin Ther 2009;31:1309–20. [55] van Wijk BL, Avorn J, Solomon DH, et al. Rates and determinants of reinitiating antihypertensive therapy after prolonged stoppage: a population-based study. J Hypertens 2007;25:689–97. [56] Frech-Tamas FH. Impact of medication burden on adherence with antihypertensive drugs [dissertation] New Brunswick, NJ: Rutgers University; 2010. [57] Yeaw J, Benner JS, Walt JG, et al. Comparing adherence and persistence across 6 chronic medication classes. J Manag Care Pharm 2009;15:728–40. [58] Friedman O, McAlister FA, Yun L, et al. Antihypertensive drug persistence and compliance among newly treated elderly hypertensives in Ontario. Am J Med 2010;123:173–81. [59] van Wijk BL, Shrank WH, Klungel OH, et al. A cross-national study of the persistence of antihypertensive medication use in the elderly. J Hypertens 2008;26:145–53. [60] Rasmussen JN, Chong A, Alter DA. Relationship between adherence to evidence-based pharmacotherapy and long-term mortality after acute myocardial infarction. JAMA 2007;297:177–86. [61] Evans CD, Eurich DT, Remillard AJ, et al. First-fill medication discontinuations and nonadherence to antihypertensive therapy: an observational study. Am J Hypertens 2012;25:195–203. [62] Gogovor A, Dragomir A, Savoie M, et al. Comparison of persistence rates with angiotensin-converting enzyme inhibitors used in secondary and primary prevention of cardiovascular disease. Value Health 2007;10:431–41. [63] Perreault S, Lamarre D, Blais L, et al. Persistence with treatment in newly treated middle-aged patients with essential hypertension. Ann Pharmacother 2005;39:1401–8.

[64] Taira DA, Gelber RP, Davis J, et al. Antihypertensive adherence and drug class among Asian Pacific Americans. Ethn Health 2007;12: 265–81. [65] Pataky RE. Persistence and adherence with cardiovascular and lipidlowering drugs following acute myocardial infarction in British Columbia. A thesis submitted to the faculty of graduate studies‐ Health Care and Epidemiology. The University of British Colombia, Vancouver. September 2009. [66] van Dijk L, Heerdink ER, Somai D, et al. Patient risk profiles and practice variation in nonadherence to antidepressants, antihypertensives and oral hypoglycemics. BMC Health Serv Res 2007;7:51. [67] Rasmussen JN, Gislason GH, Rasmussen S, et al. Use of statins and beta-blockers after acute myocardial infarction according to income and education. J Epidemiol Community Health 2007;61:1091–7. [68] Chobanian AV, Bakris GL, Black HR, et al. The seventh report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure: the JNC 7 report. JAMA 2003;289:2560–72. [69] Esposti LD, Saragoni S, Benemei S, et al. Adherence to antihypertensive medications and health outcomes among newly treated hypertensive patients. Clinicoecon Outcomes Res 2011;3:47–54. [70] Breekveldt-Postma NS, Penning-van Beest FJ, Siiskonen SJ, et al. The effect of discontinuation of antihypertensives on the risk of acute myocardial infarction and stroke. Curr Med Res Opin 2008;24:121–7. [71] Bailey JE, Wan JY, Tang J, et al. Antihypertensive medication adherence, ambulatory visits, and risk of stroke and death. J Gen Intern Med 2010;25:495–503. [72] Bouvy ML, Heerdink ER, Urquhart J, et al. Effect of a pharmacist-led intervention on diuretic compliance in heart failure patients: a randomized controlled study. J Card Fail 2003;9:404–11. [73] Crealey G, Hughes C, McElnay J, et al. Community pharmacy based provision of pharmaceutical care to older patients 2003;25:218. [74] Simpson RJ Jr, Mendys P. The effects of adherence and persistence on clinical outcomes in patients treated with statins: a systematic review. J Clin Lipidol 2010;4:462–71. [75] EN L. Compliance with lipid-lowering therapy and its impact on cardiovascular morbidity and mortality. Expert Opinion on Drug Safety 2008;7(6):717. [76] Meichenbaum D. Facilitating Treatment Adherence: A Practitioner’s Guidebook. New York: Plenum Press, 1987. [77] Balkrishnan R. Predictors of medication adherence in the elderly. Clin Ther 1998;20:764–71. [78] Ellis JJ, Erickson SR, Stevenson JG, et al. Suboptimal statin adherence and discontinuation in primary and secondary prevention populations. J Gen Intern Med 2004;19:638–45. [79] International Society for Pharmacoeconomics and Outcome Research. ASSESSING RETROSPECTIVE OBSERVATIONAL STUDIES FOR HEALTH CARE DECISIONS TASK FORCE. http://www.ispor.org/TaskForces/ InterpretingRetrospectiveObserStudiesforDMTF.asp. Updated 2013. Accessed December/05, 2013.